Move H264 dsputil functions into their own struct

This moves the H264-specific functions from DSPContext to the new
H264DSPContext.  The code is made conditional on CONFIG_H264DSP
which is set by the codecs requiring it.

The qpel and chroma MC functions are not moved as these are used by
non-h264 code.

Originally committed as revision 22565 to svn://svn.ffmpeg.org/ffmpeg/trunk

12 files changed, 663 insertions, 522 deletions

diff --git a/libavcodec/Makefile b/libavcodec/Makefile
index 66d974cdf5..84f69d802e 100644
--- a/libavcodec/Makefile
+++ b/libavcodec/Makefile
@@ -33,6 +33,7 @@ OBJS-$(CONFIG_DXVA2)                   += dxva2.o
 FFT-OBJS-$(CONFIG_HARDCODED_TABLES)    += cos_tables.o
 OBJS-$(CONFIG_FFT)                     += avfft.o fft.o $(FFT-OBJS-yes)
 OBJS-$(CONFIG_GOLOMB)                  += golomb.o
+OBJS-$(CONFIG_H264DSP)                 += h264dsp.o h264idct.o h264pred.o
 OBJS-$(CONFIG_LPC)                     += lpc.o
 OBJS-$(CONFIG_LSP)                     += lsp.o
 OBJS-$(CONFIG_MDCT)                    += mdct.o
@@ -146,7 +147,7 @@ OBJS-$(CONFIG_H263_ENCODER)            += mpegvideo_enc.o mpeg4video.o      \
                                           ratecontrol.o h263.o ituh263enc.o \
                                           flvenc.o mpeg12data.o             \
                                           mpegvideo.o error_resilience.o
-OBJS-$(CONFIG_H264_DECODER)            += h264.o h264idct.o h264pred.o         \
+OBJS-$(CONFIG_H264_DECODER)            += h264.o                               \
                                           h264_loopfilter.o h264_direct.o      \
                                           cabac.o h264_sei.o h264_ps.o         \
                                           h264_refs.o h264_cavlc.o h264_cabac.o\
@@ -282,9 +283,9 @@ OBJS-$(CONFIG_RV10_DECODER)            += rv10.o
 OBJS-$(CONFIG_RV10_ENCODER)            += rv10enc.o
 OBJS-$(CONFIG_RV20_DECODER)            += rv10.o
 OBJS-$(CONFIG_RV20_ENCODER)            += rv20enc.o
-OBJS-$(CONFIG_RV30_DECODER)            += rv30.o rv34.o h264pred.o rv30dsp.o \
+OBJS-$(CONFIG_RV30_DECODER)            += rv30.o rv34.o rv30dsp.o        \
                                           mpegvideo.o error_resilience.o
-OBJS-$(CONFIG_RV40_DECODER)            += rv40.o rv34.o h264pred.o rv40dsp.o \
+OBJS-$(CONFIG_RV40_DECODER)            += rv40.o rv34.o rv40dsp.o        \
                                           mpegvideo.o error_resilience.o
 OBJS-$(CONFIG_SGI_DECODER)             += sgidec.o
 OBJS-$(CONFIG_SGI_ENCODER)             += sgienc.o rle.o
@@ -315,7 +316,7 @@ OBJS-$(CONFIG_SVQ1_ENCODER)            += svq1enc.o svq1.o    \
                                           mpegvideo.o error_resilience.o \
                                           ituh263enc.o mpegvideo_enc.o   \
                                           ratecontrol.o mpeg12data.o
-OBJS-$(CONFIG_SVQ3_DECODER)            += h264.o svq3.o h264idct.o h264pred.o \
+OBJS-$(CONFIG_SVQ3_DECODER)            += h264.o svq3.o                       \
                                           h264_loopfilter.o h264_direct.o     \
                                           h264_sei.o h264_ps.o h264_refs.o    \
                                           h264_cavlc.o h264_cabac.o cabac.o   \
@@ -538,8 +539,8 @@ OBJS-$(CONFIG_DVBSUB_PARSER)           += dvbsub_parser.o
 OBJS-$(CONFIG_DVDSUB_PARSER)           += dvdsub_parser.o
 OBJS-$(CONFIG_H261_PARSER)             += h261_parser.o
 OBJS-$(CONFIG_H263_PARSER)             += h263_parser.o
-OBJS-$(CONFIG_H264_PARSER)             += h264_parser.o h264.o h264idct.o \
-                                          h264pred.o cabac.o              \
+OBJS-$(CONFIG_H264_PARSER)             += h264_parser.o h264.o            \
+                                          cabac.o                         \
                                           h264_refs.o h264_sei.o h264_direct.o \
                                           h264_loopfilter.o h264_cabac.o \
                                           h264_cavlc.o h264_ps.o \
@@ -631,13 +632,16 @@ OBJS-$(ARCH_ALPHA)                     += alpha/dsputil_alpha.o         \
                                           alpha/mpegvideo_alpha.o       \
                                           alpha/simple_idct_alpha.o     \
 
+ARM-OBJS-$(CONFIG_H264DSP)             += arm/h264dsp_init_arm.o        \
+                                          arm/h264pred_init_arm.o       \
+
 OBJS-$(ARCH_ARM)                       += arm/dsputil_init_arm.o        \
                                           arm/dsputil_arm.o             \
                                           arm/fft_init_arm.o            \
-                                          arm/h264pred_init_arm.o       \
                                           arm/jrevdct_arm.o             \
                                           arm/mpegvideo_arm.o           \
                                           arm/simple_idct_arm.o         \
+                                          $(ARM-OBJS-yes)
 
 OBJS-$(HAVE_ARMV5TE)                   += arm/dsputil_init_armv5te.o    \
                                           arm/mpegvideo_armv5te.o       \
@@ -658,7 +662,7 @@ NEON-OBJS-$(CONFIG_FFT)                += arm/fft_neon.o                \
 
 NEON-OBJS-$(CONFIG_MDCT)               += arm/mdct_neon.o               \
 
-NEON-OBJS-$(CONFIG_H264_DECODER)       += arm/h264dsp_neon.o            \
+NEON-OBJS-$(CONFIG_H264DSP)            += arm/h264dsp_neon.o            \
                                           arm/h264idct_neon.o           \
                                           arm/h264pred_neon.o           \
 
@@ -680,7 +684,7 @@ OBJS-$(ARCH_BFIN)                      += bfin/dsputil_bfin.o           \
 
 OBJS-$(ARCH_PPC)                       += ppc/dsputil_ppc.o             \
 
-ALTIVEC-OBJS-$(CONFIG_H264_DECODER)    += ppc/h264_altivec.o
+ALTIVEC-OBJS-$(CONFIG_H264DSP)         += ppc/h264_altivec.o
 ALTIVEC-OBJS-$(CONFIG_VC1_DECODER)     += ppc/vc1dsp_altivec.o
 ALTIVEC-OBJS-$(CONFIG_VP3_DECODER)     += ppc/vp3dsp_altivec.o
 ALTIVEC-OBJS-$(CONFIG_VP5_DECODER)     += ppc/vp3dsp_altivec.o
diff --git a/libavcodec/arm/dsputil_init_neon.c b/libavcodec/arm/dsputil_init_neon.c
index 4a8de5edc7..1f2169ead5 100644
--- a/libavcodec/arm/dsputil_init_neon.c
+++ b/libavcodec/arm/dsputil_init_neon.c
@@ -131,69 +131,6 @@ void ff_avg_h264_chroma_mc8_neon(uint8_t *, uint8_t *, int, int, int, int);
 void ff_avg_h264_chroma_mc4_neon(uint8_t *, uint8_t *, int, int, int, int);
 void ff_avg_h264_chroma_mc2_neon(uint8_t *, uint8_t *, int, int, int, int);
 
-void ff_h264_v_loop_filter_luma_neon(uint8_t *pix, int stride, int alpha,
-                                     int beta, int8_t *tc0);
-void ff_h264_h_loop_filter_luma_neon(uint8_t *pix, int stride, int alpha,
-                                     int beta, int8_t *tc0);
-void ff_h264_v_loop_filter_chroma_neon(uint8_t *pix, int stride, int alpha,
-                                       int beta, int8_t *tc0);
-void ff_h264_h_loop_filter_chroma_neon(uint8_t *pix, int stride, int alpha,
-                                       int beta, int8_t *tc0);
-
-void ff_weight_h264_pixels_16x16_neon(uint8_t *ds, int stride, int log2_den,
-                                      int weight, int offset);
-void ff_weight_h264_pixels_16x8_neon(uint8_t *ds, int stride, int log2_den,
-                                     int weight, int offset);
-void ff_weight_h264_pixels_8x16_neon(uint8_t *ds, int stride, int log2_den,
-                                     int weight, int offset);
-void ff_weight_h264_pixels_8x8_neon(uint8_t *ds, int stride, int log2_den,
-                                    int weight, int offset);
-void ff_weight_h264_pixels_8x4_neon(uint8_t *ds, int stride, int log2_den,
-                                    int weight, int offset);
-void ff_weight_h264_pixels_4x8_neon(uint8_t *ds, int stride, int log2_den,
-                                    int weight, int offset);
-void ff_weight_h264_pixels_4x4_neon(uint8_t *ds, int stride, int log2_den,
-                                    int weight, int offset);
-void ff_weight_h264_pixels_4x2_neon(uint8_t *ds, int stride, int log2_den,
-                                    int weight, int offset);
-
-void ff_biweight_h264_pixels_16x16_neon(uint8_t *dst, uint8_t *src, int stride,
-                                        int log2_den, int weightd, int weights,
-                                        int offset);
-void ff_biweight_h264_pixels_16x8_neon(uint8_t *dst, uint8_t *src, int stride,
-                                       int log2_den, int weightd, int weights,
-                                       int offset);
-void ff_biweight_h264_pixels_8x16_neon(uint8_t *dst, uint8_t *src, int stride,
-                                       int log2_den, int weightd, int weights,
-                                       int offset);
-void ff_biweight_h264_pixels_8x8_neon(uint8_t *dst, uint8_t *src, int stride,
-                                      int log2_den, int weightd, int weights,
-                                      int offset);
-void ff_biweight_h264_pixels_8x4_neon(uint8_t *dst, uint8_t *src, int stride,
-                                      int log2_den, int weightd, int weights,
-                                      int offset);
-void ff_biweight_h264_pixels_4x8_neon(uint8_t *dst, uint8_t *src, int stride,
-                                      int log2_den, int weightd, int weights,
-                                      int offset);
-void ff_biweight_h264_pixels_4x4_neon(uint8_t *dst, uint8_t *src, int stride,
-                                      int log2_den, int weightd, int weights,
-                                      int offset);
-void ff_biweight_h264_pixels_4x2_neon(uint8_t *dst, uint8_t *src, int stride,
-                                      int log2_den, int weightd, int weights,
-                                      int offset);
-
-void ff_h264_idct_add_neon(uint8_t *dst, DCTELEM *block, int stride);
-void ff_h264_idct_dc_add_neon(uint8_t *dst, DCTELEM *block, int stride);
-void ff_h264_idct_add16_neon(uint8_t *dst, const int *block_offset,
-                             DCTELEM *block, int stride,
-                             const uint8_t nnzc[6*8]);
-void ff_h264_idct_add16intra_neon(uint8_t *dst, const int *block_offset,
-                                  DCTELEM *block, int stride,
-                                  const uint8_t nnzc[6*8]);
-void ff_h264_idct_add8_neon(uint8_t **dest, const int *block_offset,
-                            DCTELEM *block, int stride,
-                            const uint8_t nnzc[6*8]);
-
 void ff_vp3_v_loop_filter_neon(uint8_t *, int, int *);
 void ff_vp3_h_loop_filter_neon(uint8_t *, int, int *);
 
@@ -352,35 +289,6 @@ void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx)
         c->avg_h264_qpel_pixels_tab[1][13] = ff_avg_h264_qpel8_mc13_neon;
         c->avg_h264_qpel_pixels_tab[1][14] = ff_avg_h264_qpel8_mc23_neon;
         c->avg_h264_qpel_pixels_tab[1][15] = ff_avg_h264_qpel8_mc33_neon;
-
-        c->h264_v_loop_filter_luma   = ff_h264_v_loop_filter_luma_neon;
-        c->h264_h_loop_filter_luma   = ff_h264_h_loop_filter_luma_neon;
-        c->h264_v_loop_filter_chroma = ff_h264_v_loop_filter_chroma_neon;
-        c->h264_h_loop_filter_chroma = ff_h264_h_loop_filter_chroma_neon;
-
-        c->weight_h264_pixels_tab[0] = ff_weight_h264_pixels_16x16_neon;
-        c->weight_h264_pixels_tab[1] = ff_weight_h264_pixels_16x8_neon;
-        c->weight_h264_pixels_tab[2] = ff_weight_h264_pixels_8x16_neon;
-        c->weight_h264_pixels_tab[3] = ff_weight_h264_pixels_8x8_neon;
-        c->weight_h264_pixels_tab[4] = ff_weight_h264_pixels_8x4_neon;
-        c->weight_h264_pixels_tab[5] = ff_weight_h264_pixels_4x8_neon;
-        c->weight_h264_pixels_tab[6] = ff_weight_h264_pixels_4x4_neon;
-        c->weight_h264_pixels_tab[7] = ff_weight_h264_pixels_4x2_neon;
-
-        c->biweight_h264_pixels_tab[0] = ff_biweight_h264_pixels_16x16_neon;
-        c->biweight_h264_pixels_tab[1] = ff_biweight_h264_pixels_16x8_neon;
-        c->biweight_h264_pixels_tab[2] = ff_biweight_h264_pixels_8x16_neon;
-        c->biweight_h264_pixels_tab[3] = ff_biweight_h264_pixels_8x8_neon;
-        c->biweight_h264_pixels_tab[4] = ff_biweight_h264_pixels_8x4_neon;
-        c->biweight_h264_pixels_tab[5] = ff_biweight_h264_pixels_4x8_neon;
-        c->biweight_h264_pixels_tab[6] = ff_biweight_h264_pixels_4x4_neon;
-        c->biweight_h264_pixels_tab[7] = ff_biweight_h264_pixels_4x2_neon;
-
-        c->h264_idct_add        = ff_h264_idct_add_neon;
-        c->h264_idct_dc_add     = ff_h264_idct_dc_add_neon;
-        c->h264_idct_add16      = ff_h264_idct_add16_neon;
-        c->h264_idct_add16intra = ff_h264_idct_add16intra_neon;
-        c->h264_idct_add8       = ff_h264_idct_add8_neon;
     }
 
     if (CONFIG_VP3_DECODER) {
diff --git a/libavcodec/arm/h264dsp_init_arm.c b/libavcodec/arm/h264dsp_init_arm.c
new file mode 100644
index 0000000000..c06c3d0c3d
--- /dev/null
+++ b/libavcodec/arm/h264dsp_init_arm.c
@@ -0,0 +1,126 @@
+/*
+ * Copyright (c) 2010 Mans Rullgard <mans@mansr.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdint.h>
+
+#include "libavcodec/dsputil.h"
+#include "libavcodec/h264dsp.h"
+
+void ff_h264_v_loop_filter_luma_neon(uint8_t *pix, int stride, int alpha,
+                                     int beta, int8_t *tc0);
+void ff_h264_h_loop_filter_luma_neon(uint8_t *pix, int stride, int alpha,
+                                     int beta, int8_t *tc0);
+void ff_h264_v_loop_filter_chroma_neon(uint8_t *pix, int stride, int alpha,
+                                       int beta, int8_t *tc0);
+void ff_h264_h_loop_filter_chroma_neon(uint8_t *pix, int stride, int alpha,
+                                       int beta, int8_t *tc0);
+
+void ff_weight_h264_pixels_16x16_neon(uint8_t *ds, int stride, int log2_den,
+                                      int weight, int offset);
+void ff_weight_h264_pixels_16x8_neon(uint8_t *ds, int stride, int log2_den,
+                                     int weight, int offset);
+void ff_weight_h264_pixels_8x16_neon(uint8_t *ds, int stride, int log2_den,
+                                     int weight, int offset);
+void ff_weight_h264_pixels_8x8_neon(uint8_t *ds, int stride, int log2_den,
+                                    int weight, int offset);
+void ff_weight_h264_pixels_8x4_neon(uint8_t *ds, int stride, int log2_den,
+                                    int weight, int offset);
+void ff_weight_h264_pixels_4x8_neon(uint8_t *ds, int stride, int log2_den,
+                                    int weight, int offset);
+void ff_weight_h264_pixels_4x4_neon(uint8_t *ds, int stride, int log2_den,
+                                    int weight, int offset);
+void ff_weight_h264_pixels_4x2_neon(uint8_t *ds, int stride, int log2_den,
+                                    int weight, int offset);
+
+void ff_biweight_h264_pixels_16x16_neon(uint8_t *dst, uint8_t *src, int stride,
+                                        int log2_den, int weightd, int weights,
+                                        int offset);
+void ff_biweight_h264_pixels_16x8_neon(uint8_t *dst, uint8_t *src, int stride,
+                                       int log2_den, int weightd, int weights,
+                                       int offset);
+void ff_biweight_h264_pixels_8x16_neon(uint8_t *dst, uint8_t *src, int stride,
+                                       int log2_den, int weightd, int weights,
+                                       int offset);
+void ff_biweight_h264_pixels_8x8_neon(uint8_t *dst, uint8_t *src, int stride,
+                                      int log2_den, int weightd, int weights,
+                                      int offset);
+void ff_biweight_h264_pixels_8x4_neon(uint8_t *dst, uint8_t *src, int stride,
+                                      int log2_den, int weightd, int weights,
+                                      int offset);
+void ff_biweight_h264_pixels_4x8_neon(uint8_t *dst, uint8_t *src, int stride,
+                                      int log2_den, int weightd, int weights,
+                                      int offset);
+void ff_biweight_h264_pixels_4x4_neon(uint8_t *dst, uint8_t *src, int stride,
+                                      int log2_den, int weightd, int weights,
+                                      int offset);
+void ff_biweight_h264_pixels_4x2_neon(uint8_t *dst, uint8_t *src, int stride,
+                                      int log2_den, int weightd, int weights,
+                                      int offset);
+
+void ff_h264_idct_add_neon(uint8_t *dst, DCTELEM *block, int stride);
+void ff_h264_idct_dc_add_neon(uint8_t *dst, DCTELEM *block, int stride);
+void ff_h264_idct_add16_neon(uint8_t *dst, const int *block_offset,
+                             DCTELEM *block, int stride,
+                             const uint8_t nnzc[6*8]);
+void ff_h264_idct_add16intra_neon(uint8_t *dst, const int *block_offset,
+                                  DCTELEM *block, int stride,
+                                  const uint8_t nnzc[6*8]);
+void ff_h264_idct_add8_neon(uint8_t **dest, const int *block_offset,
+                            DCTELEM *block, int stride,
+                            const uint8_t nnzc[6*8]);
+
+#if HAVE_NEON
+static void ff_h264dsp_init_neon(H264DSPContext *c)
+{
+    c->h264_v_loop_filter_luma   = ff_h264_v_loop_filter_luma_neon;
+    c->h264_h_loop_filter_luma   = ff_h264_h_loop_filter_luma_neon;
+    c->h264_v_loop_filter_chroma = ff_h264_v_loop_filter_chroma_neon;
+    c->h264_h_loop_filter_chroma = ff_h264_h_loop_filter_chroma_neon;
+
+    c->weight_h264_pixels_tab[0] = ff_weight_h264_pixels_16x16_neon;
+    c->weight_h264_pixels_tab[1] = ff_weight_h264_pixels_16x8_neon;
+    c->weight_h264_pixels_tab[2] = ff_weight_h264_pixels_8x16_neon;
+    c->weight_h264_pixels_tab[3] = ff_weight_h264_pixels_8x8_neon;
+    c->weight_h264_pixels_tab[4] = ff_weight_h264_pixels_8x4_neon;
+    c->weight_h264_pixels_tab[5] = ff_weight_h264_pixels_4x8_neon;
+    c->weight_h264_pixels_tab[6] = ff_weight_h264_pixels_4x4_neon;
+    c->weight_h264_pixels_tab[7] = ff_weight_h264_pixels_4x2_neon;
+
+    c->biweight_h264_pixels_tab[0] = ff_biweight_h264_pixels_16x16_neon;
+    c->biweight_h264_pixels_tab[1] = ff_biweight_h264_pixels_16x8_neon;
+    c->biweight_h264_pixels_tab[2] = ff_biweight_h264_pixels_8x16_neon;
+    c->biweight_h264_pixels_tab[3] = ff_biweight_h264_pixels_8x8_neon;
+    c->biweight_h264_pixels_tab[4] = ff_biweight_h264_pixels_8x4_neon;
+    c->biweight_h264_pixels_tab[5] = ff_biweight_h264_pixels_4x8_neon;
+    c->biweight_h264_pixels_tab[6] = ff_biweight_h264_pixels_4x4_neon;
+    c->biweight_h264_pixels_tab[7] = ff_biweight_h264_pixels_4x2_neon;
+
+    c->h264_idct_add        = ff_h264_idct_add_neon;
+    c->h264_idct_dc_add     = ff_h264_idct_dc_add_neon;
+    c->h264_idct_add16      = ff_h264_idct_add16_neon;
+    c->h264_idct_add16intra = ff_h264_idct_add16intra_neon;
+    c->h264_idct_add8       = ff_h264_idct_add8_neon;
+}
+#endif
+
+void ff_h264dsp_init_arm(H264DSPContext *c)
+{
+    if (HAVE_NEON) ff_h264dsp_init_neon(c);
+}
diff --git a/libavcodec/dsputil.c b/libavcodec/dsputil.c
index 8535c6d665..bbfdb6ae8d 100644
--- a/libavcodec/dsputil.c
+++ b/libavcodec/dsputil.c
@@ -2597,76 +2597,6 @@ H264_MC(avg_, 16)
 #undef op2_put
 #endif
 
-#define op_scale1(x)  block[x] = av_clip_uint8( (block[x]*weight + offset) >> log2_denom )
-#define op_scale2(x)  dst[x] = av_clip_uint8( (src[x]*weights + dst[x]*weightd + offset) >> (log2_denom+1))
-#define H264_WEIGHT(W,H) \
-static void weight_h264_pixels ## W ## x ## H ## _c(uint8_t *block, int stride, int log2_denom, int weight, int offset){ \
-    int y; \
-    offset <<= log2_denom; \
-    if(log2_denom) offset += 1<<(log2_denom-1); \
-    for(y=0; y<H; y++, block += stride){ \
-        op_scale1(0); \
-        op_scale1(1); \
-        if(W==2) continue; \
-        op_scale1(2); \
-        op_scale1(3); \
-        if(W==4) continue; \
-        op_scale1(4); \
-        op_scale1(5); \
-        op_scale1(6); \
-        op_scale1(7); \
-        if(W==8) continue; \
-        op_scale1(8); \
-        op_scale1(9); \
-        op_scale1(10); \
-        op_scale1(11); \
-        op_scale1(12); \
-        op_scale1(13); \
-        op_scale1(14); \
-        op_scale1(15); \
-    } \
-} \
-static void biweight_h264_pixels ## W ## x ## H ## _c(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offset){ \
-    int y; \
-    offset = ((offset + 1) | 1) << log2_denom; \
-    for(y=0; y<H; y++, dst += stride, src += stride){ \
-        op_scale2(0); \
-        op_scale2(1); \
-        if(W==2) continue; \
-        op_scale2(2); \
-        op_scale2(3); \
-        if(W==4) continue; \
-        op_scale2(4); \
-        op_scale2(5); \
-        op_scale2(6); \
-        op_scale2(7); \
-        if(W==8) continue; \
-        op_scale2(8); \
-        op_scale2(9); \
-        op_scale2(10); \
-        op_scale2(11); \
-        op_scale2(12); \
-        op_scale2(13); \
-        op_scale2(14); \
-        op_scale2(15); \
-    } \
-}
-
-H264_WEIGHT(16,16)
-H264_WEIGHT(16,8)
-H264_WEIGHT(8,16)
-H264_WEIGHT(8,8)
-H264_WEIGHT(8,4)
-H264_WEIGHT(4,8)
-H264_WEIGHT(4,4)
-H264_WEIGHT(4,2)
-H264_WEIGHT(2,4)
-H264_WEIGHT(2,2)
-
-#undef op_scale1
-#undef op_scale2
-#undef H264_WEIGHT
-
 static void wmv2_mspel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){
     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
     int i;
@@ -2711,9 +2641,6 @@ void ff_avg_vc1_mspel_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int r
 }
 #endif /* CONFIG_VC1_DECODER */
 
-/* H264 specific */
-void ff_h264dspenc_init(DSPContext* c, AVCodecContext *avctx);
-
 #if CONFIG_RV40_DECODER
 static void put_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride){
     put_pixels16_xy2_c(dst, src, stride, 16);
@@ -2907,179 +2834,6 @@ static void h261_loop_filter_c(uint8_t *src, int stride){
     }
 }
 
-static av_always_inline av_flatten void h264_loop_filter_luma_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0)
-{
-    int i, d;
-    for( i = 0; i < 4; i++ ) {
-        if( tc0[i] < 0 ) {
-            pix += 4*ystride;
-            continue;
-        }
-        for( d = 0; d < 4; d++ ) {
-            const int p0 = pix[-1*xstride];
-            const int p1 = pix[-2*xstride];
-            const int p2 = pix[-3*xstride];
-            const int q0 = pix[0];
-            const int q1 = pix[1*xstride];
-            const int q2 = pix[2*xstride];
-
-            if( FFABS( p0 - q0 ) < alpha &&
-                FFABS( p1 - p0 ) < beta &&
-                FFABS( q1 - q0 ) < beta ) {
-
-                int tc = tc0[i];
-                int i_delta;
-
-                if( FFABS( p2 - p0 ) < beta ) {
-                    if(tc0[i])
-                    pix[-2*xstride] = p1 + av_clip( (( p2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - p1, -tc0[i], tc0[i] );
-                    tc++;
-                }
-                if( FFABS( q2 - q0 ) < beta ) {
-                    if(tc0[i])
-                    pix[   xstride] = q1 + av_clip( (( q2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - q1, -tc0[i], tc0[i] );
-                    tc++;
-                }
-
-                i_delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
-                pix[-xstride] = av_clip_uint8( p0 + i_delta );    /* p0' */
-                pix[0]        = av_clip_uint8( q0 - i_delta );    /* q0' */
-            }
-            pix += ystride;
-        }
-    }
-}
-static void h264_v_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
-{
-    h264_loop_filter_luma_c(pix, stride, 1, alpha, beta, tc0);
-}
-static void h264_h_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
-{
-    h264_loop_filter_luma_c(pix, 1, stride, alpha, beta, tc0);
-}
-
-static av_always_inline av_flatten void h264_loop_filter_luma_intra_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta)
-{
-    int d;
-    for( d = 0; d < 16; d++ ) {
-        const int p2 = pix[-3*xstride];
-        const int p1 = pix[-2*xstride];
-        const int p0 = pix[-1*xstride];
-
-        const int q0 = pix[ 0*xstride];
-        const int q1 = pix[ 1*xstride];
-        const int q2 = pix[ 2*xstride];
-
-        if( FFABS( p0 - q0 ) < alpha &&
-            FFABS( p1 - p0 ) < beta &&
-            FFABS( q1 - q0 ) < beta ) {
-
-            if(FFABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){
-                if( FFABS( p2 - p0 ) < beta)
-                {
-                    const int p3 = pix[-4*xstride];
-                    /* p0', p1', p2' */
-                    pix[-1*xstride] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
-                    pix[-2*xstride] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
-                    pix[-3*xstride] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
-                } else {
-                    /* p0' */
-                    pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
-                }
-                if( FFABS( q2 - q0 ) < beta)
-                {
-                    const int q3 = pix[3*xstride];
-                    /* q0', q1', q2' */
-                    pix[0*xstride] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
-                    pix[1*xstride] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
-                    pix[2*xstride] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
-                } else {
-                    /* q0' */
-                    pix[0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
-                }
-            }else{
-                /* p0', q0' */
-                pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
-                pix[ 0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
-            }
-        }
-        pix += ystride;
-    }
-}
-static void h264_v_loop_filter_luma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
-{
-    h264_loop_filter_luma_intra_c(pix, stride, 1, alpha, beta);
-}
-static void h264_h_loop_filter_luma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
-{
-    h264_loop_filter_luma_intra_c(pix, 1, stride, alpha, beta);
-}
-
-static av_always_inline av_flatten void h264_loop_filter_chroma_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0)
-{
-    int i, d;
-    for( i = 0; i < 4; i++ ) {
-        const int tc = tc0[i];
-        if( tc <= 0 ) {
-            pix += 2*ystride;
-            continue;
-        }
-        for( d = 0; d < 2; d++ ) {
-            const int p0 = pix[-1*xstride];
-            const int p1 = pix[-2*xstride];
-            const int q0 = pix[0];
-            const int q1 = pix[1*xstride];
-
-            if( FFABS( p0 - q0 ) < alpha &&
-                FFABS( p1 - p0 ) < beta &&
-                FFABS( q1 - q0 ) < beta ) {
-
-                int delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
-
-                pix[-xstride] = av_clip_uint8( p0 + delta );    /* p0' */
-                pix[0]        = av_clip_uint8( q0 - delta );    /* q0' */
-            }
-            pix += ystride;
-        }
-    }
-}
-static void h264_v_loop_filter_chroma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
-{
-    h264_loop_filter_chroma_c(pix, stride, 1, alpha, beta, tc0);
-}
-static void h264_h_loop_filter_chroma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
-{
-    h264_loop_filter_chroma_c(pix, 1, stride, alpha, beta, tc0);
-}
-
-static av_always_inline av_flatten void h264_loop_filter_chroma_intra_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta)
-{
-    int d;
-    for( d = 0; d < 8; d++ ) {
-        const int p0 = pix[-1*xstride];
-        const int p1 = pix[-2*xstride];
-        const int q0 = pix[0];
-        const int q1 = pix[1*xstride];
-
-        if( FFABS( p0 - q0 ) < alpha &&
-            FFABS( p1 - p0 ) < beta &&
-            FFABS( q1 - q0 ) < beta ) {
-
-            pix[-xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;   /* p0' */
-            pix[0]        = ( 2*q1 + q0 + p1 + 2 ) >> 2;   /* q0' */
-        }
-        pix += ystride;
-    }
-}
-static void h264_v_loop_filter_chroma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
-{
-    h264_loop_filter_chroma_intra_c(pix, stride, 1, alpha, beta);
-}
-static void h264_h_loop_filter_chroma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
-{
-    h264_loop_filter_chroma_intra_c(pix, 1, stride, alpha, beta);
-}
-
 static inline int pix_abs16_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
 {
     int s, i;
@@ -4502,17 +4256,6 @@ av_cold void dsputil_init(DSPContext* c, AVCodecContext *avctx)
         }
     }
 
-    if (CONFIG_H264_DECODER) {
-        c->h264_idct_add= ff_h264_idct_add_c;
-        c->h264_idct8_add= ff_h264_idct8_add_c;
-        c->h264_idct_dc_add= ff_h264_idct_dc_add_c;
-        c->h264_idct8_dc_add= ff_h264_idct8_dc_add_c;
-        c->h264_idct_add16     = ff_h264_idct_add16_c;
-        c->h264_idct8_add4     = ff_h264_idct8_add4_c;
-        c->h264_idct_add8      = ff_h264_idct_add8_c;
-        c->h264_idct_add16intra= ff_h264_idct_add16intra_c;
-    }
-
     c->get_pixels = get_pixels_c;
     c->diff_pixels = diff_pixels_c;
     c->put_pixels_clamped = put_pixels_clamped_c;
@@ -4635,27 +4378,6 @@ av_cold void dsputil_init(DSPContext* c, AVCodecContext *avctx)
     c->put_no_rnd_vc1_chroma_pixels_tab[0]= put_no_rnd_vc1_chroma_mc8_c;
     c->avg_no_rnd_vc1_chroma_pixels_tab[0]= avg_no_rnd_vc1_chroma_mc8_c;
 
-    c->weight_h264_pixels_tab[0]= weight_h264_pixels16x16_c;
-    c->weight_h264_pixels_tab[1]= weight_h264_pixels16x8_c;
-    c->weight_h264_pixels_tab[2]= weight_h264_pixels8x16_c;
-    c->weight_h264_pixels_tab[3]= weight_h264_pixels8x8_c;
-    c->weight_h264_pixels_tab[4]= weight_h264_pixels8x4_c;
-    c->weight_h264_pixels_tab[5]= weight_h264_pixels4x8_c;
-    c->weight_h264_pixels_tab[6]= weight_h264_pixels4x4_c;
-    c->weight_h264_pixels_tab[7]= weight_h264_pixels4x2_c;
-    c->weight_h264_pixels_tab[8]= weight_h264_pixels2x4_c;
-    c->weight_h264_pixels_tab[9]= weight_h264_pixels2x2_c;
-    c->biweight_h264_pixels_tab[0]= biweight_h264_pixels16x16_c;
-    c->biweight_h264_pixels_tab[1]= biweight_h264_pixels16x8_c;
-    c->biweight_h264_pixels_tab[2]= biweight_h264_pixels8x16_c;
-    c->biweight_h264_pixels_tab[3]= biweight_h264_pixels8x8_c;
-    c->biweight_h264_pixels_tab[4]= biweight_h264_pixels8x4_c;
-    c->biweight_h264_pixels_tab[5]= biweight_h264_pixels4x8_c;
-    c->biweight_h264_pixels_tab[6]= biweight_h264_pixels4x4_c;
-    c->biweight_h264_pixels_tab[7]= biweight_h264_pixels4x2_c;
-    c->biweight_h264_pixels_tab[8]= biweight_h264_pixels2x4_c;
-    c->biweight_h264_pixels_tab[9]= biweight_h264_pixels2x2_c;
-
     c->draw_edges = draw_edges_c;
 
 #if CONFIG_CAVS_DECODER
@@ -4737,16 +4459,6 @@ av_cold void dsputil_init(DSPContext* c, AVCodecContext *avctx)
     c->add_png_paeth_prediction= ff_add_png_paeth_prediction;
 #endif
 
-    c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_c;
-    c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_c;
-    c->h264_v_loop_filter_luma_intra= h264_v_loop_filter_luma_intra_c;
-    c->h264_h_loop_filter_luma_intra= h264_h_loop_filter_luma_intra_c;
-    c->h264_v_loop_filter_chroma= h264_v_loop_filter_chroma_c;
-    c->h264_h_loop_filter_chroma= h264_h_loop_filter_chroma_c;
-    c->h264_v_loop_filter_chroma_intra= h264_v_loop_filter_chroma_intra_c;
-    c->h264_h_loop_filter_chroma_intra= h264_h_loop_filter_chroma_intra_c;
-    c->h264_loop_filter_strength= NULL;
-
     if (CONFIG_H263_DECODER || CONFIG_H263_ENCODER) {
         c->h263_h_loop_filter= h263_h_loop_filter_c;
         c->h263_v_loop_filter= h263_v_loop_filter_c;
diff --git a/libavcodec/dsputil.h b/libavcodec/dsputil.h
index 5a2246cbee..d1816e66ba 100644
--- a/libavcodec/dsputil.h
+++ b/libavcodec/dsputil.h
@@ -149,8 +149,6 @@ typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const ui
 typedef void (*tpel_mc_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int w, int h);
 typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
 typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y);
-typedef void (*h264_weight_func)(uint8_t *block, int stride, int log2_denom, int weight, int offset);
-typedef void (*h264_biweight_func)(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offset);
 
 typedef void (*op_fill_func)(uint8_t *block/*align width (8 or 16)*/, uint8_t value, int line_size, int h);
 
@@ -340,9 +338,6 @@ typedef struct DSPContext {
     qpel_mc_func put_2tap_qpel_pixels_tab[4][16];
     qpel_mc_func avg_2tap_qpel_pixels_tab[4][16];
 
-    h264_weight_func weight_h264_pixels_tab[10];
-    h264_biweight_func biweight_h264_pixels_tab[10];
-
     /* AVS specific */
     qpel_mc_func put_cavs_qpel_pixels_tab[2][16];
     qpel_mc_func avg_cavs_qpel_pixels_tab[2][16];
@@ -370,19 +365,6 @@ typedef struct DSPContext {
     void (*add_png_paeth_prediction)(uint8_t *dst, uint8_t *src, uint8_t *top, int w, int bpp);
     void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w);
 
-    void (*h264_v_loop_filter_luma)(uint8_t *pix/*align 16*/, int stride, int alpha, int beta, int8_t *tc0);
-    void (*h264_h_loop_filter_luma)(uint8_t *pix/*align 4 */, int stride, int alpha, int beta, int8_t *tc0);
-    /* v/h_loop_filter_luma_intra: align 16 */
-    void (*h264_v_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta);
-    void (*h264_h_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta);
-    void (*h264_v_loop_filter_chroma)(uint8_t *pix/*align 8*/, int stride, int alpha, int beta, int8_t *tc0);
-    void (*h264_h_loop_filter_chroma)(uint8_t *pix/*align 4*/, int stride, int alpha, int beta, int8_t *tc0);
-    void (*h264_v_loop_filter_chroma_intra)(uint8_t *pix/*align 8*/, int stride, int alpha, int beta);
-    void (*h264_h_loop_filter_chroma_intra)(uint8_t *pix/*align 8*/, int stride, int alpha, int beta);
-    // h264_loop_filter_strength: simd only. the C version is inlined in h264.c
-    void (*h264_loop_filter_strength)(int16_t bS[2][4][4], uint8_t nnz[40], int8_t ref[2][40], int16_t mv[2][40][2],
-                                      int bidir, int edges, int step, int mask_mv0, int mask_mv1, int field);
-
     void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
     void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);
 
@@ -517,21 +499,6 @@ typedef struct DSPContext {
     void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w);
 #define EDGE_WIDTH 16
 
-    /* h264 functions */
-    /* NOTE!!! if you implement any of h264_idct8_add, h264_idct8_add4 then you must implement all of them
-       NOTE!!! if you implement any of h264_idct_add, h264_idct_add16, h264_idct_add16intra, h264_idct_add8 then you must implement all of them
-        The reason for above, is that no 2 out of one list may use a different permutation.
-    */
-    void (*h264_idct_add)(uint8_t *dst/*align 4*/, DCTELEM *block/*align 16*/, int stride);
-    void (*h264_idct8_add)(uint8_t *dst/*align 8*/, DCTELEM *block/*align 16*/, int stride);
-    void (*h264_idct_dc_add)(uint8_t *dst/*align 4*/, DCTELEM *block/*align 16*/, int stride);
-    void (*h264_idct8_dc_add)(uint8_t *dst/*align 8*/, DCTELEM *block/*align 16*/, int stride);
-    void (*h264_dct)(DCTELEM block[4][4]);
-    void (*h264_idct_add16)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
-    void (*h264_idct8_add4)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
-    void (*h264_idct_add8)(uint8_t **dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
-    void (*h264_idct_add16intra)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
-
     void (*prefetch)(void *mem, int stride, int h);
 
     void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
diff --git a/libavcodec/h264.c b/libavcodec/h264.c
index a177117ae0..9fe7c22208 100644
--- a/libavcodec/h264.c
+++ b/libavcodec/h264.c
@@ -678,7 +678,7 @@ static void free_tables(H264Context *h){
 
 static void init_dequant8_coeff_table(H264Context *h){
     int i,q,x;
-    const int transpose = (h->s.dsp.h264_idct8_add != ff_h264_idct8_add_c); //FIXME ugly
+    const int transpose = (h->h264dsp.h264_idct8_add != ff_h264_idct8_add_c); //FIXME ugly
     h->dequant8_coeff[0] = h->dequant8_buffer[0];
     h->dequant8_coeff[1] = h->dequant8_buffer[1];
 
@@ -701,7 +701,7 @@ static void init_dequant8_coeff_table(H264Context *h){
 
 static void init_dequant4_coeff_table(H264Context *h){
     int i,j,q,x;
-    const int transpose = (h->s.dsp.h264_idct_add != ff_h264_idct_add_c); //FIXME ugly
+    const int transpose = (h->h264dsp.h264_idct_add != ff_h264_idct_add_c); //FIXME ugly
     for(i=0; i<6; i++ ){
         h->dequant4_coeff[i] = h->dequant4_buffer[i];
         for(j=0; j<i; j++){
@@ -831,6 +831,7 @@ static av_cold void common_init(H264Context *h){
     s->height = s->avctx->height;
     s->codec_id= s->avctx->codec->id;
 
+    ff_h264dsp_init(&h->h264dsp);
     ff_h264_pred_init(&h->hpc, s->codec_id);
 
     h->dequant_coeff_pps= -1;
@@ -1159,8 +1160,8 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple){
                             idct_dc_add =
                             idct_add    = s->dsp.add_pixels8;
                         }else{
-                            idct_dc_add = s->dsp.h264_idct8_dc_add;
-                            idct_add    = s->dsp.h264_idct8_add;
+                            idct_dc_add = h->h264dsp.h264_idct8_dc_add;
+                            idct_add    = h->h264dsp.h264_idct8_add;
                         }
                         for(i=0; i<16; i+=4){
                             uint8_t * const ptr= dest_y + block_offset[i];
@@ -1184,8 +1185,8 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple){
                             idct_dc_add =
                             idct_add    = s->dsp.add_pixels4;
                         }else{
-                            idct_dc_add = s->dsp.h264_idct_dc_add;
-                            idct_add    = s->dsp.h264_idct_add;
+                            idct_dc_add = h->h264dsp.h264_idct_dc_add;
+                            idct_add    = h->h264dsp.h264_idct_add;
                         }
                         for(i=0; i<16; i++){
                             uint8_t * const ptr= dest_y + block_offset[i];
@@ -1236,7 +1237,7 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple){
             hl_motion(h, dest_y, dest_cb, dest_cr,
                       s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab,
                       s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab,
-                      s->dsp.weight_h264_pixels_tab, s->dsp.biweight_h264_pixels_tab);
+                      h->h264dsp.weight_h264_pixels_tab, h->h264dsp.biweight_h264_pixels_tab);
         }
 
 
@@ -1253,7 +1254,7 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple){
                             }
                         }
                     }else{
-                         s->dsp.h264_idct_add16intra(dest_y, block_offset, h->mb, linesize, h->non_zero_count_cache);
+                         h->h264dsp.h264_idct_add16intra(dest_y, block_offset, h->mb, linesize, h->non_zero_count_cache);
                     }
                 }else if(h->cbp&15){
                     if(transform_bypass){
@@ -1266,9 +1267,9 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple){
                         }
                     }else{
                         if(IS_8x8DCT(mb_type)){
-                            s->dsp.h264_idct8_add4(dest_y, block_offset, h->mb, linesize, h->non_zero_count_cache);
+                            h->h264dsp.h264_idct8_add4(dest_y, block_offset, h->mb, linesize, h->non_zero_count_cache);
                         }else{
-                            s->dsp.h264_idct_add16(dest_y, block_offset, h->mb, linesize, h->non_zero_count_cache);
+                            h->h264dsp.h264_idct_add16(dest_y, block_offset, h->mb, linesize, h->non_zero_count_cache);
                         }
                     }
                 }
@@ -1299,8 +1300,8 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple){
                 chroma_dc_dequant_idct_c(h->mb + 16*16, h->chroma_qp[0], h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][h->chroma_qp[0]][0]);
                 chroma_dc_dequant_idct_c(h->mb + 16*16+4*16, h->chroma_qp[1], h->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][h->chroma_qp[1]][0]);
                 if(is_h264){
-                    idct_add = s->dsp.h264_idct_add;
-                    idct_dc_add = s->dsp.h264_idct_dc_add;
+                    idct_add = h->h264dsp.h264_idct_add;
+                    idct_dc_add = h->h264dsp.h264_idct_dc_add;
                     for(i=16; i<16+8; i++){
                         if(h->non_zero_count_cache[ scan8[i] ])
                             idct_add   (dest[(i&4)>>2] + block_offset[i], h->mb + i*16, uvlinesize);
@@ -1560,7 +1561,7 @@ static int init_poc(H264Context *h){
 static void init_scan_tables(H264Context *h){
     MpegEncContext * const s = &h->s;
     int i;
-    if(s->dsp.h264_idct_add == ff_h264_idct_add_c){ //FIXME little ugly
+    if(h->h264dsp.h264_idct_add == ff_h264_idct_add_c){ //FIXME little ugly
         memcpy(h->zigzag_scan, zigzag_scan, 16*sizeof(uint8_t));
         memcpy(h-> field_scan,  field_scan, 16*sizeof(uint8_t));
     }else{
@@ -1571,7 +1572,7 @@ static void init_scan_tables(H264Context *h){
 #undef T
         }
     }
-    if(s->dsp.h264_idct8_add == ff_h264_idct8_add_c){
+    if(h->h264dsp.h264_idct8_add == ff_h264_idct8_add_c){
         memcpy(h->zigzag_scan8x8,       ff_zigzag_direct,     64*sizeof(uint8_t));
         memcpy(h->zigzag_scan8x8_cavlc, zigzag_scan8x8_cavlc, 64*sizeof(uint8_t));
         memcpy(h->field_scan8x8,        field_scan8x8,        64*sizeof(uint8_t));
@@ -3003,7 +3004,7 @@ int main(void){
         }
 //        printf("\n");
 
-        s->dsp.h264_idct_add(ref, block, 4);
+        h->h264dsp.h264_idct_add(ref, block, 4);
 /*        for(j=0; j<16; j++){
             printf("%d ", ref[j]);
         }
diff --git a/libavcodec/h264.h b/libavcodec/h264.h
index 479ade9074..965e39d9bd 100644
--- a/libavcodec/h264.h
+++ b/libavcodec/h264.h
@@ -32,6 +32,7 @@
 #include "dsputil.h"
 #include "cabac.h"
 #include "mpegvideo.h"
+#include "h264dsp.h"
 #include "h264pred.h"
 #include "rectangle.h"
 
@@ -262,6 +263,7 @@ typedef struct MMCO{
  */
 typedef struct H264Context{
     MpegEncContext s;
+    H264DSPContext h264dsp;
     int chroma_qp[2]; //QPc
 
     int qp_thresh;      ///< QP threshold to skip loopfilter
diff --git a/libavcodec/h264_loopfilter.c b/libavcodec/h264_loopfilter.c
index 0fca3d4562..a3696b3d5c 100644
--- a/libavcodec/h264_loopfilter.c
+++ b/libavcodec/h264_loopfilter.c
@@ -112,9 +112,9 @@ static void av_always_inline filter_mb_edgev( uint8_t *pix, int stride, int16_t
         tc[1] = tc0_table[index_a][bS[1]];
         tc[2] = tc0_table[index_a][bS[2]];
         tc[3] = tc0_table[index_a][bS[3]];
-        h->s.dsp.h264_h_loop_filter_luma(pix, stride, alpha, beta, tc);
+        h->h264dsp.h264_h_loop_filter_luma(pix, stride, alpha, beta, tc);
     } else {
-        h->s.dsp.h264_h_loop_filter_luma_intra(pix, stride, alpha, beta);
+        h->h264dsp.h264_h_loop_filter_luma_intra(pix, stride, alpha, beta);
     }
 }
 static void av_always_inline filter_mb_edgecv( uint8_t *pix, int stride, int16_t bS[4], unsigned int qp, H264Context *h ) {
@@ -129,9 +129,9 @@ static void av_always_inline filter_mb_edgecv( uint8_t *pix, int stride, int16_t
         tc[1] = tc0_table[index_a][bS[1]]+1;
         tc[2] = tc0_table[index_a][bS[2]]+1;
         tc[3] = tc0_table[index_a][bS[3]]+1;
-        h->s.dsp.h264_h_loop_filter_chroma(pix, stride, alpha, beta, tc);
+        h->h264dsp.h264_h_loop_filter_chroma(pix, stride, alpha, beta, tc);
     } else {
-        h->s.dsp.h264_h_loop_filter_chroma_intra(pix, stride, alpha, beta);
+        h->h264dsp.h264_h_loop_filter_chroma_intra(pix, stride, alpha, beta);
     }
 }
 
@@ -282,9 +282,9 @@ static void av_always_inline filter_mb_edgeh( uint8_t *pix, int stride, int16_t
         tc[1] = tc0_table[index_a][bS[1]];
         tc[2] = tc0_table[index_a][bS[2]];
         tc[3] = tc0_table[index_a][bS[3]];
-        h->s.dsp.h264_v_loop_filter_luma(pix, stride, alpha, beta, tc);
+        h->h264dsp.h264_v_loop_filter_luma(pix, stride, alpha, beta, tc);
     } else {
-        h->s.dsp.h264_v_loop_filter_luma_intra(pix, stride, alpha, beta);
+        h->h264dsp.h264_v_loop_filter_luma_intra(pix, stride, alpha, beta);
     }
 }
 
@@ -300,9 +300,9 @@ static void av_always_inline filter_mb_edgech( uint8_t *pix, int stride, int16_t
         tc[1] = tc0_table[index_a][bS[1]]+1;
         tc[2] = tc0_table[index_a][bS[2]]+1;
         tc[3] = tc0_table[index_a][bS[3]]+1;
-        h->s.dsp.h264_v_loop_filter_chroma(pix, stride, alpha, beta, tc);
+        h->h264dsp.h264_v_loop_filter_chroma(pix, stride, alpha, beta, tc);
     } else {
-        h->s.dsp.h264_v_loop_filter_chroma_intra(pix, stride, alpha, beta);
+        h->h264dsp.h264_v_loop_filter_chroma_intra(pix, stride, alpha, beta);
     }
 }
 
@@ -314,7 +314,7 @@ void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y,
 
     mb_xy = h->mb_xy;
 
-    if(!h->top_type || !s->dsp.h264_loop_filter_strength || h->pps.chroma_qp_diff) {
+    if(!h->top_type || !h->h264dsp.h264_loop_filter_strength || h->pps.chroma_qp_diff) {
         ff_h264_filter_mb(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize);
         return;
     }
@@ -381,7 +381,7 @@ void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y,
             int mask_edge0 = 3*((mask_edge1>>1) & ((5*left_type)>>5)&1); // (mb_type & (MB_TYPE_16x16 | MB_TYPE_8x16)) && (h->left_type[0] & (MB_TYPE_16x16 | MB_TYPE_8x16)) ? 3 : 0;
             int step =  1+(mb_type>>24); //IS_8x8DCT(mb_type) ? 2 : 1;
             edges = 4 - 3*((mb_type>>3) & !(h->cbp & 15)); //(mb_type & MB_TYPE_16x16) && !(h->cbp & 15) ? 1 : 4;
-            s->dsp.h264_loop_filter_strength( bS, h->non_zero_count_cache, h->ref_cache, h->mv_cache,
+            h->h264dsp.h264_loop_filter_strength( bS, h->non_zero_count_cache, h->ref_cache, h->mv_cache,
                                               h->list_count==2, edges, step, mask_edge0, mask_edge1, FIELD_PICTURE);
         }
         if( IS_INTRA(left_type) )
diff --git a/libavcodec/h264dsp.c b/libavcodec/h264dsp.c
new file mode 100644
index 0000000000..d65df1b1ed
--- /dev/null
+++ b/libavcodec/h264dsp.c
@@ -0,0 +1,320 @@
+/*
+ * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
+ * Copyright (c) 2003-2010 Michael Niedermayer <michaelni@gmx.at>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file libavcodec/h264dsp.c
+ * H.264 / AVC / MPEG4 part10 DSP functions.
+ * @author Michael Niedermayer <michaelni@gmx.at>
+ */
+
+#include <stdint.h>
+#include "avcodec.h"
+#include "h264dsp.h"
+
+#define op_scale1(x)  block[x] = av_clip_uint8( (block[x]*weight + offset) >> log2_denom )
+#define op_scale2(x)  dst[x] = av_clip_uint8( (src[x]*weights + dst[x]*weightd + offset) >> (log2_denom+1))
+#define H264_WEIGHT(W,H) \
+static void weight_h264_pixels ## W ## x ## H ## _c(uint8_t *block, int stride, int log2_denom, int weight, int offset){ \
+    int y; \
+    offset <<= log2_denom; \
+    if(log2_denom) offset += 1<<(log2_denom-1); \
+    for(y=0; y<H; y++, block += stride){ \
+        op_scale1(0); \
+        op_scale1(1); \
+        if(W==2) continue; \
+        op_scale1(2); \
+        op_scale1(3); \
+        if(W==4) continue; \
+        op_scale1(4); \
+        op_scale1(5); \
+        op_scale1(6); \
+        op_scale1(7); \
+        if(W==8) continue; \
+        op_scale1(8); \
+        op_scale1(9); \
+        op_scale1(10); \
+        op_scale1(11); \
+        op_scale1(12); \
+        op_scale1(13); \
+        op_scale1(14); \
+        op_scale1(15); \
+    } \
+} \
+static void biweight_h264_pixels ## W ## x ## H ## _c(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offset){ \
+    int y; \
+    offset = ((offset + 1) | 1) << log2_denom; \
+    for(y=0; y<H; y++, dst += stride, src += stride){ \
+        op_scale2(0); \
+        op_scale2(1); \
+        if(W==2) continue; \
+        op_scale2(2); \
+        op_scale2(3); \
+        if(W==4) continue; \
+        op_scale2(4); \
+        op_scale2(5); \
+        op_scale2(6); \
+        op_scale2(7); \
+        if(W==8) continue; \
+        op_scale2(8); \
+        op_scale2(9); \
+        op_scale2(10); \
+        op_scale2(11); \
+        op_scale2(12); \
+        op_scale2(13); \
+        op_scale2(14); \
+        op_scale2(15); \
+    } \
+}
+
+H264_WEIGHT(16,16)
+H264_WEIGHT(16,8)
+H264_WEIGHT(8,16)
+H264_WEIGHT(8,8)
+H264_WEIGHT(8,4)
+H264_WEIGHT(4,8)
+H264_WEIGHT(4,4)
+H264_WEIGHT(4,2)
+H264_WEIGHT(2,4)
+H264_WEIGHT(2,2)
+
+#undef op_scale1
+#undef op_scale2
+#undef H264_WEIGHT
+
+static av_always_inline av_flatten void h264_loop_filter_luma_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0)
+{
+    int i, d;
+    for( i = 0; i < 4; i++ ) {
+        if( tc0[i] < 0 ) {
+            pix += 4*ystride;
+            continue;
+        }
+        for( d = 0; d < 4; d++ ) {
+            const int p0 = pix[-1*xstride];
+            const int p1 = pix[-2*xstride];
+            const int p2 = pix[-3*xstride];
+            const int q0 = pix[0];
+            const int q1 = pix[1*xstride];
+            const int q2 = pix[2*xstride];
+
+            if( FFABS( p0 - q0 ) < alpha &&
+                FFABS( p1 - p0 ) < beta &&
+                FFABS( q1 - q0 ) < beta ) {
+
+                int tc = tc0[i];
+                int i_delta;
+
+                if( FFABS( p2 - p0 ) < beta ) {
+                    if(tc0[i])
+                    pix[-2*xstride] = p1 + av_clip( (( p2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - p1, -tc0[i], tc0[i] );
+                    tc++;
+                }
+                if( FFABS( q2 - q0 ) < beta ) {
+                    if(tc0[i])
+                    pix[   xstride] = q1 + av_clip( (( q2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - q1, -tc0[i], tc0[i] );
+                    tc++;
+                }
+
+                i_delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
+                pix[-xstride] = av_clip_uint8( p0 + i_delta );    /* p0' */
+                pix[0]        = av_clip_uint8( q0 - i_delta );    /* q0' */
+            }
+            pix += ystride;
+        }
+    }
+}
+static void h264_v_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
+{
+    h264_loop_filter_luma_c(pix, stride, 1, alpha, beta, tc0);
+}
+static void h264_h_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
+{
+    h264_loop_filter_luma_c(pix, 1, stride, alpha, beta, tc0);
+}
+
+static av_always_inline av_flatten void h264_loop_filter_luma_intra_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta)
+{
+    int d;
+    for( d = 0; d < 16; d++ ) {
+        const int p2 = pix[-3*xstride];
+        const int p1 = pix[-2*xstride];
+        const int p0 = pix[-1*xstride];
+
+        const int q0 = pix[ 0*xstride];
+        const int q1 = pix[ 1*xstride];
+        const int q2 = pix[ 2*xstride];
+
+        if( FFABS( p0 - q0 ) < alpha &&
+            FFABS( p1 - p0 ) < beta &&
+            FFABS( q1 - q0 ) < beta ) {
+
+            if(FFABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){
+                if( FFABS( p2 - p0 ) < beta)
+                {
+                    const int p3 = pix[-4*xstride];
+                    /* p0', p1', p2' */
+                    pix[-1*xstride] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
+                    pix[-2*xstride] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
+                    pix[-3*xstride] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
+                } else {
+                    /* p0' */
+                    pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
+                }
+                if( FFABS( q2 - q0 ) < beta)
+                {
+                    const int q3 = pix[3*xstride];
+                    /* q0', q1', q2' */
+                    pix[0*xstride] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
+                    pix[1*xstride] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
+                    pix[2*xstride] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
+                } else {
+                    /* q0' */
+                    pix[0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
+                }
+            }else{
+                /* p0', q0' */
+                pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
+                pix[ 0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
+            }
+        }
+        pix += ystride;
+    }
+}
+static void h264_v_loop_filter_luma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
+{
+    h264_loop_filter_luma_intra_c(pix, stride, 1, alpha, beta);
+}
+static void h264_h_loop_filter_luma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
+{
+    h264_loop_filter_luma_intra_c(pix, 1, stride, alpha, beta);
+}
+
+static av_always_inline av_flatten void h264_loop_filter_chroma_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0)
+{
+    int i, d;
+    for( i = 0; i < 4; i++ ) {
+        const int tc = tc0[i];
+        if( tc <= 0 ) {
+            pix += 2*ystride;
+            continue;
+        }
+        for( d = 0; d < 2; d++ ) {
+            const int p0 = pix[-1*xstride];
+            const int p1 = pix[-2*xstride];
+            const int q0 = pix[0];
+            const int q1 = pix[1*xstride];
+
+            if( FFABS( p0 - q0 ) < alpha &&
+                FFABS( p1 - p0 ) < beta &&
+                FFABS( q1 - q0 ) < beta ) {
+
+                int delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
+
+                pix[-xstride] = av_clip_uint8( p0 + delta );    /* p0' */
+                pix[0]        = av_clip_uint8( q0 - delta );    /* q0' */
+            }
+            pix += ystride;
+        }
+    }
+}
+static void h264_v_loop_filter_chroma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
+{
+    h264_loop_filter_chroma_c(pix, stride, 1, alpha, beta, tc0);
+}
+static void h264_h_loop_filter_chroma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
+{
+    h264_loop_filter_chroma_c(pix, 1, stride, alpha, beta, tc0);
+}
+
+static av_always_inline av_flatten void h264_loop_filter_chroma_intra_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta)
+{
+    int d;
+    for( d = 0; d < 8; d++ ) {
+        const int p0 = pix[-1*xstride];
+        const int p1 = pix[-2*xstride];
+        const int q0 = pix[0];
+        const int q1 = pix[1*xstride];
+
+        if( FFABS( p0 - q0 ) < alpha &&
+            FFABS( p1 - p0 ) < beta &&
+            FFABS( q1 - q0 ) < beta ) {
+
+            pix[-xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;   /* p0' */
+            pix[0]        = ( 2*q1 + q0 + p1 + 2 ) >> 2;   /* q0' */
+        }
+        pix += ystride;
+    }
+}
+static void h264_v_loop_filter_chroma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
+{
+    h264_loop_filter_chroma_intra_c(pix, stride, 1, alpha, beta);
+}
+static void h264_h_loop_filter_chroma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
+{
+    h264_loop_filter_chroma_intra_c(pix, 1, stride, alpha, beta);
+}
+
+void ff_h264dsp_init(H264DSPContext *c)
+{
+    c->h264_idct_add= ff_h264_idct_add_c;
+    c->h264_idct8_add= ff_h264_idct8_add_c;
+    c->h264_idct_dc_add= ff_h264_idct_dc_add_c;
+    c->h264_idct8_dc_add= ff_h264_idct8_dc_add_c;
+    c->h264_idct_add16     = ff_h264_idct_add16_c;
+    c->h264_idct8_add4     = ff_h264_idct8_add4_c;
+    c->h264_idct_add8      = ff_h264_idct_add8_c;
+    c->h264_idct_add16intra= ff_h264_idct_add16intra_c;
+
+    c->weight_h264_pixels_tab[0]= weight_h264_pixels16x16_c;
+    c->weight_h264_pixels_tab[1]= weight_h264_pixels16x8_c;
+    c->weight_h264_pixels_tab[2]= weight_h264_pixels8x16_c;
+    c->weight_h264_pixels_tab[3]= weight_h264_pixels8x8_c;
+    c->weight_h264_pixels_tab[4]= weight_h264_pixels8x4_c;
+    c->weight_h264_pixels_tab[5]= weight_h264_pixels4x8_c;
+    c->weight_h264_pixels_tab[6]= weight_h264_pixels4x4_c;
+    c->weight_h264_pixels_tab[7]= weight_h264_pixels4x2_c;
+    c->weight_h264_pixels_tab[8]= weight_h264_pixels2x4_c;
+    c->weight_h264_pixels_tab[9]= weight_h264_pixels2x2_c;
+    c->biweight_h264_pixels_tab[0]= biweight_h264_pixels16x16_c;
+    c->biweight_h264_pixels_tab[1]= biweight_h264_pixels16x8_c;
+    c->biweight_h264_pixels_tab[2]= biweight_h264_pixels8x16_c;
+    c->biweight_h264_pixels_tab[3]= biweight_h264_pixels8x8_c;
+    c->biweight_h264_pixels_tab[4]= biweight_h264_pixels8x4_c;
+    c->biweight_h264_pixels_tab[5]= biweight_h264_pixels4x8_c;
+    c->biweight_h264_pixels_tab[6]= biweight_h264_pixels4x4_c;
+    c->biweight_h264_pixels_tab[7]= biweight_h264_pixels4x2_c;
+    c->biweight_h264_pixels_tab[8]= biweight_h264_pixels2x4_c;
+    c->biweight_h264_pixels_tab[9]= biweight_h264_pixels2x2_c;
+
+    c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_c;
+    c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_c;
+    c->h264_v_loop_filter_luma_intra= h264_v_loop_filter_luma_intra_c;
+    c->h264_h_loop_filter_luma_intra= h264_h_loop_filter_luma_intra_c;
+    c->h264_v_loop_filter_chroma= h264_v_loop_filter_chroma_c;
+    c->h264_h_loop_filter_chroma= h264_h_loop_filter_chroma_c;
+    c->h264_v_loop_filter_chroma_intra= h264_v_loop_filter_chroma_intra_c;
+    c->h264_h_loop_filter_chroma_intra= h264_h_loop_filter_chroma_intra_c;
+    c->h264_loop_filter_strength= NULL;
+
+    if (ARCH_ARM) ff_h264dsp_init_arm(c);
+    if (ARCH_PPC) ff_h264dsp_init_ppc(c);
+    if (HAVE_MMX) ff_h264dsp_init_x86(c);
+}
diff --git a/libavcodec/h264dsp.h b/libavcodec/h264dsp.h
new file mode 100644
index 0000000000..454f434c99
--- /dev/null
+++ b/libavcodec/h264dsp.h
@@ -0,0 +1,80 @@
+/*
+ * Copyright (c) 2003-2010 Michael Niedermayer <michaelni@gmx.at>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file libavcodec/h264dsp.h
+ * H.264 DSP functions.
+ * @author Michael Niedermayer <michaelni@gmx.at>
+ */
+
+#ifndef AVCODEC_H264DSP_H
+#define AVCODEC_H264DSP_H
+
+#include <stdint.h>
+#include "dsputil.h"
+
+//typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y);
+typedef void (*h264_weight_func)(uint8_t *block, int stride, int log2_denom, int weight, int offset);
+typedef void (*h264_biweight_func)(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offset);
+
+/**
+ * Context for storing H.264 DSP functions
+ */
+typedef struct H264DSPContext{
+    /* weighted MC */
+    h264_weight_func weight_h264_pixels_tab[10];
+    h264_biweight_func biweight_h264_pixels_tab[10];
+
+    /* loop filter */
+    void (*h264_v_loop_filter_luma)(uint8_t *pix/*align 16*/, int stride, int alpha, int beta, int8_t *tc0);
+    void (*h264_h_loop_filter_luma)(uint8_t *pix/*align 4 */, int stride, int alpha, int beta, int8_t *tc0);
+    /* v/h_loop_filter_luma_intra: align 16 */
+    void (*h264_v_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta);
+    void (*h264_h_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta);
+    void (*h264_v_loop_filter_chroma)(uint8_t *pix/*align 8*/, int stride, int alpha, int beta, int8_t *tc0);
+    void (*h264_h_loop_filter_chroma)(uint8_t *pix/*align 4*/, int stride, int alpha, int beta, int8_t *tc0);
+    void (*h264_v_loop_filter_chroma_intra)(uint8_t *pix/*align 8*/, int stride, int alpha, int beta);
+    void (*h264_h_loop_filter_chroma_intra)(uint8_t *pix/*align 8*/, int stride, int alpha, int beta);
+    // h264_loop_filter_strength: simd only. the C version is inlined in h264.c
+    void (*h264_loop_filter_strength)(int16_t bS[2][4][4], uint8_t nnz[40], int8_t ref[2][40], int16_t mv[2][40][2],
+                                      int bidir, int edges, int step, int mask_mv0, int mask_mv1, int field);
+
+    /* IDCT */
+    /* NOTE!!! if you implement any of h264_idct8_add, h264_idct8_add4 then you must implement all of them
+       NOTE!!! if you implement any of h264_idct_add, h264_idct_add16, h264_idct_add16intra, h264_idct_add8 then you must implement all of them
+        The reason for above, is that no 2 out of one list may use a different permutation.
+    */
+    void (*h264_idct_add)(uint8_t *dst/*align 4*/, DCTELEM *block/*align 16*/, int stride);
+    void (*h264_idct8_add)(uint8_t *dst/*align 8*/, DCTELEM *block/*align 16*/, int stride);
+    void (*h264_idct_dc_add)(uint8_t *dst/*align 4*/, DCTELEM *block/*align 16*/, int stride);
+    void (*h264_idct8_dc_add)(uint8_t *dst/*align 8*/, DCTELEM *block/*align 16*/, int stride);
+    void (*h264_dct)(DCTELEM block[4][4]);
+    void (*h264_idct_add16)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
+    void (*h264_idct8_add4)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
+    void (*h264_idct_add8)(uint8_t **dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
+    void (*h264_idct_add16intra)(uint8_t *dst/*align 16*/, const int *blockoffset, DCTELEM *block/*align 16*/, int stride, const uint8_t nnzc[6*8]);
+}H264DSPContext;
+
+void ff_h264dsp_init(H264DSPContext *c);
+void ff_h264dsp_init_arm(H264DSPContext *c);
+void ff_h264dsp_init_ppc(H264DSPContext *c);
+void ff_h264dsp_init_x86(H264DSPContext *c);
+
+#endif /* AVCODEC_H264DSP_H */
diff --git a/libavcodec/ppc/h264_altivec.c b/libavcodec/ppc/h264_altivec.c
index d6329882d1..47e416b82b 100644
--- a/libavcodec/ppc/h264_altivec.c
+++ b/libavcodec/ppc/h264_altivec.c
@@ -20,6 +20,7 @@
 
 #include "libavcodec/dsputil.h"
 #include "libavcodec/h264data.h"
+#include "libavcodec/h264dsp.h"
 
 #include "dsputil_ppc.h"
 #include "dsputil_altivec.h"
@@ -974,16 +975,6 @@ void dsputil_h264_init_ppc(DSPContext* c, AVCodecContext *avctx) {
         c->avg_h264_chroma_pixels_tab[0] = avg_h264_chroma_mc8_altivec;
         c->put_no_rnd_vc1_chroma_pixels_tab[0] = put_no_rnd_vc1_chroma_mc8_altivec;
         c->avg_no_rnd_vc1_chroma_pixels_tab[0] = avg_no_rnd_vc1_chroma_mc8_altivec;
-        c->h264_idct_add = ff_h264_idct_add_altivec;
-        c->h264_idct_add8 = ff_h264_idct_add8_altivec;
-        c->h264_idct_add16 = ff_h264_idct_add16_altivec;
-        c->h264_idct_add16intra = ff_h264_idct_add16intra_altivec;
-        c->h264_idct_dc_add= h264_idct_dc_add_altivec;
-        c->h264_idct8_dc_add = ff_h264_idct8_dc_add_altivec;
-        c->h264_idct8_add = ff_h264_idct8_add_altivec;
-        c->h264_idct8_add4 = ff_h264_idct8_add4_altivec;
-        c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_altivec;
-        c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_altivec;
 
 #define dspfunc(PFX, IDX, NUM) \
         c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_altivec; \
@@ -1006,6 +997,22 @@ void dsputil_h264_init_ppc(DSPContext* c, AVCodecContext *avctx) {
         dspfunc(put_h264_qpel, 0, 16);
         dspfunc(avg_h264_qpel, 0, 16);
 #undef dspfunc
+    }
+}
+
+void ff_h264dsp_init_ppc(H264DSPContext *c)
+{
+    if (has_altivec()) {
+        c->h264_idct_add = ff_h264_idct_add_altivec;
+        c->h264_idct_add8 = ff_h264_idct_add8_altivec;
+        c->h264_idct_add16 = ff_h264_idct_add16_altivec;
+        c->h264_idct_add16intra = ff_h264_idct_add16intra_altivec;
+        c->h264_idct_dc_add= h264_idct_dc_add_altivec;
+        c->h264_idct8_dc_add = ff_h264_idct8_dc_add_altivec;
+        c->h264_idct8_add = ff_h264_idct8_add_altivec;
+        c->h264_idct8_add4 = ff_h264_idct8_add4_altivec;
+        c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_altivec;
+        c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_altivec;
 
         c->weight_h264_pixels_tab[0] = ff_weight_h264_pixels16x16_altivec;
         c->weight_h264_pixels_tab[1] = ff_weight_h264_pixels16x8_altivec;
diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c
index 216cb4318d..31387ebb30 100644
--- a/libavcodec/x86/dsputil_mmx.c
+++ b/libavcodec/x86/dsputil_mmx.c
@@ -24,6 +24,7 @@
 
 #include "libavutil/x86_cpu.h"
 #include "libavcodec/dsputil.h"
+#include "libavcodec/h264dsp.h"
 #include "libavcodec/mpegvideo.h"
 #include "libavcodec/simple_idct.h"
 #include "dsputil_mmx.h"
@@ -2618,16 +2619,6 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
         c->put_rv40_chroma_pixels_tab[0]= put_rv40_chroma_mc8_mmx;
         c->put_rv40_chroma_pixels_tab[1]= put_rv40_chroma_mc4_mmx;
 
-        c->h264_idct_dc_add=
-        c->h264_idct_add= ff_h264_idct_add_mmx;
-        c->h264_idct8_dc_add=
-        c->h264_idct8_add= ff_h264_idct8_add_mmx;
-
-        c->h264_idct_add16     = ff_h264_idct_add16_mmx;
-        c->h264_idct8_add4     = ff_h264_idct8_add4_mmx;
-        c->h264_idct_add8      = ff_h264_idct_add8_mmx;
-        c->h264_idct_add16intra= ff_h264_idct_add16intra_mmx;
-
         if (CONFIG_VP6_DECODER) {
             c->vp6_filter_diag4 = ff_vp6_filter_diag4_mmx;
         }
@@ -2649,13 +2640,6 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
             c->avg_pixels_tab[1][1] = avg_pixels8_x2_mmx2;
             c->avg_pixels_tab[1][2] = avg_pixels8_y2_mmx2;
 
-            c->h264_idct_dc_add= ff_h264_idct_dc_add_mmx2;
-            c->h264_idct8_dc_add= ff_h264_idct8_dc_add_mmx2;
-            c->h264_idct_add16     = ff_h264_idct_add16_mmx2;
-            c->h264_idct8_add4     = ff_h264_idct8_add4_mmx2;
-            c->h264_idct_add8      = ff_h264_idct_add8_mmx2;
-            c->h264_idct_add16intra= ff_h264_idct_add16intra_mmx2;
-
             if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
                 c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_mmx2;
                 c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_mmx2;
@@ -2716,31 +2700,6 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
             c->avg_h264_chroma_pixels_tab[1]= avg_h264_chroma_mc4_mmx2;
             c->avg_h264_chroma_pixels_tab[2]= avg_h264_chroma_mc2_mmx2;
             c->put_h264_chroma_pixels_tab[2]= put_h264_chroma_mc2_mmx2;
-            c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_mmx2;
-            c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_mmx2;
-            c->h264_v_loop_filter_chroma= h264_v_loop_filter_chroma_mmx2;
-            c->h264_h_loop_filter_chroma= h264_h_loop_filter_chroma_mmx2;
-            c->h264_v_loop_filter_chroma_intra= h264_v_loop_filter_chroma_intra_mmx2;
-            c->h264_h_loop_filter_chroma_intra= h264_h_loop_filter_chroma_intra_mmx2;
-            c->h264_loop_filter_strength= h264_loop_filter_strength_mmx2;
-
-            c->weight_h264_pixels_tab[0]= ff_h264_weight_16x16_mmx2;
-            c->weight_h264_pixels_tab[1]= ff_h264_weight_16x8_mmx2;
-            c->weight_h264_pixels_tab[2]= ff_h264_weight_8x16_mmx2;
-            c->weight_h264_pixels_tab[3]= ff_h264_weight_8x8_mmx2;
-            c->weight_h264_pixels_tab[4]= ff_h264_weight_8x4_mmx2;
-            c->weight_h264_pixels_tab[5]= ff_h264_weight_4x8_mmx2;
-            c->weight_h264_pixels_tab[6]= ff_h264_weight_4x4_mmx2;
-            c->weight_h264_pixels_tab[7]= ff_h264_weight_4x2_mmx2;
-
-            c->biweight_h264_pixels_tab[0]= ff_h264_biweight_16x16_mmx2;
-            c->biweight_h264_pixels_tab[1]= ff_h264_biweight_16x8_mmx2;
-            c->biweight_h264_pixels_tab[2]= ff_h264_biweight_8x16_mmx2;
-            c->biweight_h264_pixels_tab[3]= ff_h264_biweight_8x8_mmx2;
-            c->biweight_h264_pixels_tab[4]= ff_h264_biweight_8x4_mmx2;
-            c->biweight_h264_pixels_tab[5]= ff_h264_biweight_4x8_mmx2;
-            c->biweight_h264_pixels_tab[6]= ff_h264_biweight_4x4_mmx2;
-            c->biweight_h264_pixels_tab[7]= ff_h264_biweight_4x2_mmx2;
 
 #if HAVE_YASM
             c->add_hfyu_median_prediction = ff_add_hfyu_median_prediction_mmx2;
@@ -2825,9 +2784,6 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
             H264_QPEL_FUNCS(0, 0, sse2);
         }
         if(mm_flags & FF_MM_SSE2){
-            c->h264_idct8_add = ff_h264_idct8_add_sse2;
-            c->h264_idct8_add4= ff_h264_idct8_add4_sse2;
-
             H264_QPEL_FUNCS(0, 1, sse2);
             H264_QPEL_FUNCS(0, 2, sse2);
             H264_QPEL_FUNCS(0, 3, sse2);
@@ -2874,26 +2830,6 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
         }
 #endif
 
-#if CONFIG_GPL && HAVE_YASM
-        if (mm_flags & FF_MM_MMX2){
-#if ARCH_X86_32
-            c->h264_v_loop_filter_luma_intra = ff_x264_deblock_v_luma_intra_mmxext;
-            c->h264_h_loop_filter_luma_intra = ff_x264_deblock_h_luma_intra_mmxext;
-#endif
-            if( mm_flags&FF_MM_SSE2 ){
-#if ARCH_X86_64 || !defined(__ICC) || __ICC > 1110
-                c->h264_v_loop_filter_luma = ff_x264_deblock_v_luma_sse2;
-                c->h264_h_loop_filter_luma = ff_x264_deblock_h_luma_sse2;
-                c->h264_v_loop_filter_luma_intra = ff_x264_deblock_v_luma_intra_sse2;
-                c->h264_h_loop_filter_luma_intra = ff_x264_deblock_h_luma_intra_sse2;
-#endif
-                c->h264_idct_add16 = ff_h264_idct_add16_sse2;
-                c->h264_idct_add8  = ff_h264_idct_add8_sse2;
-                c->h264_idct_add16intra = ff_h264_idct_add16intra_sse2;
-            }
-        }
-#endif
-
         if(mm_flags & FF_MM_3DNOW){
             c->vorbis_inverse_coupling = vorbis_inverse_coupling_3dnow;
             c->vector_fmul = vector_fmul_3dnow;
@@ -2983,3 +2919,81 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
     //ff_idct = just_return;
 #endif
 }
+
+#if CONFIG_H264DSP
+void ff_h264dsp_init_x86(H264DSPContext *c)
+{
+    mm_flags = mm_support();
+
+    if (mm_flags & FF_MM_MMX) {
+        c->h264_idct_dc_add=
+        c->h264_idct_add= ff_h264_idct_add_mmx;
+        c->h264_idct8_dc_add=
+        c->h264_idct8_add= ff_h264_idct8_add_mmx;
+
+        c->h264_idct_add16     = ff_h264_idct_add16_mmx;
+        c->h264_idct8_add4     = ff_h264_idct8_add4_mmx;
+        c->h264_idct_add8      = ff_h264_idct_add8_mmx;
+        c->h264_idct_add16intra= ff_h264_idct_add16intra_mmx;
+
+        if (mm_flags & FF_MM_MMX2) {
+            c->h264_idct_dc_add= ff_h264_idct_dc_add_mmx2;
+            c->h264_idct8_dc_add= ff_h264_idct8_dc_add_mmx2;
+            c->h264_idct_add16     = ff_h264_idct_add16_mmx2;
+            c->h264_idct8_add4     = ff_h264_idct8_add4_mmx2;
+            c->h264_idct_add8      = ff_h264_idct_add8_mmx2;
+            c->h264_idct_add16intra= ff_h264_idct_add16intra_mmx2;
+
+            c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_mmx2;
+            c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_mmx2;
+            c->h264_v_loop_filter_chroma= h264_v_loop_filter_chroma_mmx2;
+            c->h264_h_loop_filter_chroma= h264_h_loop_filter_chroma_mmx2;
+            c->h264_v_loop_filter_chroma_intra= h264_v_loop_filter_chroma_intra_mmx2;
+            c->h264_h_loop_filter_chroma_intra= h264_h_loop_filter_chroma_intra_mmx2;
+            c->h264_loop_filter_strength= h264_loop_filter_strength_mmx2;
+
+            c->weight_h264_pixels_tab[0]= ff_h264_weight_16x16_mmx2;
+            c->weight_h264_pixels_tab[1]= ff_h264_weight_16x8_mmx2;
+            c->weight_h264_pixels_tab[2]= ff_h264_weight_8x16_mmx2;
+            c->weight_h264_pixels_tab[3]= ff_h264_weight_8x8_mmx2;
+            c->weight_h264_pixels_tab[4]= ff_h264_weight_8x4_mmx2;
+            c->weight_h264_pixels_tab[5]= ff_h264_weight_4x8_mmx2;
+            c->weight_h264_pixels_tab[6]= ff_h264_weight_4x4_mmx2;
+            c->weight_h264_pixels_tab[7]= ff_h264_weight_4x2_mmx2;
+
+            c->biweight_h264_pixels_tab[0]= ff_h264_biweight_16x16_mmx2;
+            c->biweight_h264_pixels_tab[1]= ff_h264_biweight_16x8_mmx2;
+            c->biweight_h264_pixels_tab[2]= ff_h264_biweight_8x16_mmx2;
+            c->biweight_h264_pixels_tab[3]= ff_h264_biweight_8x8_mmx2;
+            c->biweight_h264_pixels_tab[4]= ff_h264_biweight_8x4_mmx2;
+            c->biweight_h264_pixels_tab[5]= ff_h264_biweight_4x8_mmx2;
+            c->biweight_h264_pixels_tab[6]= ff_h264_biweight_4x4_mmx2;
+            c->biweight_h264_pixels_tab[7]= ff_h264_biweight_4x2_mmx2;
+        }
+        if(mm_flags & FF_MM_SSE2){
+            c->h264_idct8_add = ff_h264_idct8_add_sse2;
+            c->h264_idct8_add4= ff_h264_idct8_add4_sse2;
+        }
+
+#if CONFIG_GPL && HAVE_YASM
+        if (mm_flags & FF_MM_MMX2){
+#if ARCH_X86_32
+            c->h264_v_loop_filter_luma_intra = ff_x264_deblock_v_luma_intra_mmxext;
+            c->h264_h_loop_filter_luma_intra = ff_x264_deblock_h_luma_intra_mmxext;
+#endif
+            if( mm_flags&FF_MM_SSE2 ){
+#if ARCH_X86_64 || !defined(__ICC) || __ICC > 1110
+                c->h264_v_loop_filter_luma = ff_x264_deblock_v_luma_sse2;
+                c->h264_h_loop_filter_luma = ff_x264_deblock_h_luma_sse2;
+                c->h264_v_loop_filter_luma_intra = ff_x264_deblock_v_luma_intra_sse2;
+                c->h264_h_loop_filter_luma_intra = ff_x264_deblock_h_luma_intra_sse2;
+#endif
+                c->h264_idct_add16 = ff_h264_idct_add16_sse2;
+                c->h264_idct_add8  = ff_h264_idct_add8_sse2;
+                c->h264_idct_add16intra = ff_h264_idct_add16intra_sse2;
+            }
+        }
+#endif
+    }
+}
+#endif /* CONFIG_H264DSP */